Robert Blumofe

Executive VP & CTO, Akamai

Search every verified Robert Blumofe interview, podcast appearance, and on-the-record quote — each transcript cross-checked by AI and human review to confirm speaker identity. Robert Blumofe, Executive Vice President and CTO of Akamai, published a tutorial on March 21, 2026, demonstrating a physical analog implementation of a perceptron, which he described as a single-neuron neural network. In the video, Blumofe explained that the device provides a hands-on illustration of key concepts in deep learning, including training, inference, and generalization. He credited a video by Welch Labs for the original idea and build plan, noting that he modified the design by using a printed circuit board to simplify assembly. Blumofe provided a build guide and encouraged viewers to construct their own perceptron. During the tutorial, Blumofe demonstrated how the perceptron could be trained using sample inputs and labels, adjusting weight knobs to produce correct outputs for patterns representing the letters T and J. He discussed the importance of training samples and the process of iterating through them to achieve generalization, where a neural net gives correct outputs for inputs not seen during training. Blumofe remarked that while the perceptron could be trained with each sample used once, larger neural networks typically require multiple passes through training data.